IEEE Transactions on Learning Technologies

Recommender systems have been researched extensively by the Technology Enhanced Learning (TEL) community during the last decade. By identifying suitable resources from a potentially overwhelming variety of choices, such systems offer a promising approach to facilitate both learning and teaching tasks. As learning is taking place in extremely diverse and rich environments, the incorporation of contextual information about the user in the recommendation process has attracted major interest. Such contextualization is researched as a paradigm for building intelligent systems that can better predict and anticipate the needs of users, and act more efficiently in response to their behavior. In this paper, we try to assess the degree to which current work in TEL recommender systems has achieved this, as well as outline areas in which further work is needed. First, we present a context framework that identifies relevant context dimensions for TEL applications. Then, we present an analysis of existing TEL recommender systems along these dimensions. Finally, based on our survey results, we outline topics on which further research is needed.

This paper presents a systematic literature review of the state-of-the-art of research on learning dashboards in the fields of Learning Analytics and Educational Data Mining. Research on learning dashboards aims to identify what data is meaningful to different stakeholders and how data can be presented to support sense-making processes. Learning dashboards are becoming popular due to the increased use of educational technologies, such as Learning Management Systems (LMS) and Massive Open Online Courses (MOOCs). The initial search of five main academic databases and GScholar resulted in 346 papers out of which 55 papers were included in the final analysis. Our review distinguishes different kinds of research studies as well as various aspects of learning dashboards and their maturity regarding evaluation. As the research field is still relatively young, most studies are exploratory and proof-of-concept. The review concludes by offering a definition for learning dashboards and by outlining open issues and future lines of work in the area of learning dashboards. There is a need for longitudinal research in authentic settings and studies that systematically compare different dashboard designs.

Gamification is the use of game design elements in non-game settings to engage participants and encourage desired behaviors. It has been identified as a promising technique to improve students' engagement which could have a positive impact on learning. This study evaluated the learning effectiveness and engagement appeal of a gamified learning activity targeted at the learning of C-programming language. Furthermore, the study inquired into which gamified learning activities were more appealing to students. The study was conducted using the mixed-method sequential explanatory protocol. The data collected and analysed included logs, questionnaires, and pre- and post-tests. The results of the evaluation show positive effects on the engagement of students toward the gamified learning activities and a moderate improvement in learning outcomes. Students reported different motivations for continuing and stopping activities once they completed the mandatory assignment. The preferences for different gamified activities were also conditioned by academic milestones.

Augmented reality (AR) technology is mature for creating learning experiences for K-12 (pre-school, grade school, and high school) educational settings. We reviewed the applications intended to complement traditional curriculum materials for K-12. We found 87 research articles on augmented reality learning experiences (ARLEs) in the IEEE Xplore Digital Library and other learning technology publications. Forty-three of these articles conducted user studies, and seven allowed the computation of an effect size to the performance of students in a test. In our meta-analysis, research show that ARLEs achieved a widely variable effect on student performance from a small negative effect to a large effect, with a mean effect size of 0.56 or moderate effect. To complement this finding, we performed a qualitative analysis on the design aspects for ARLEs: display hardware, software libraries, content authoring solutions, and evaluation techniques. We explain that AR incur three inherent advantages: real world annotation, contextual visualization, and vision-haptic visualization. We illustrate these advantages through the exemplifying prototypes, and ground these advantages to multimedia learning theory, experiential learning theory, and animate vision theory. Insights from this review are aimed to inform the design of future ARLEs.

In this paper, a new dance training system based on the motion capture and virtual reality (VR) technologies is proposed. Our system is inspired by the traditional way to learn new movements-imitating the teacher's movements and listening to the teacher's feedback. A prototype of our proposed system is implemented, in which a student can imitate the motion demonstrated by a virtual teacher projected on the wall screen. Meanwhile, the student's motions will be captured and analyzed by the system based on which feedback is given back to them. The result of user studies showed that our system can successfully guide students to improve their skills. The subjects agreed that the system is interesting and can motivate them to learn.

This paper presents a systematic literature review of learning analytics dashboards (LADs) research that reports empirical findings to assess the impact on learning and teaching. Several previous literature reviews identified self-regulated learning as a primary focus of LADs. However, there has been much less understanding how learning analytics are grounded in the literature on self-regulated learning and how self-regulated learning is supported. To address this limitation, this review analyzed the existing empirical studies on LADs based on the well-known model of self-regulated learning proposed by Winne and Hadwin. The results show that existing LADs are rarely grounded in learning theory, cannot be suggested to support metacognition, do not offer any information about effective learning tactics and strategies, and have significant limitations in how their evaluation is conducted and reported. Based on the findings of the study and through the synthesis of the literature, the paper proposes that future research and development should not make any a priori design decisions about representation of data and analytic results in learning analytics systems such as LADs. To formalize this proposal, the paper defines the model for user-centered learning analytics systems (MULAS). The MULAS consists of the four dimensions that are cyclically and recursively interconnected including: theory, design, feedback, and evaluation.

As the fields of learning analytics and learning design mature, the convergence and synergies between the two are becoming an important area for research. This paper intends to summarize the main outcomes of a systematic review of empirical evidence on learning analytics for learning design. Moreover, this paper presents an overview of what and how learning analytics have been used to inform learning design decisions and in what contexts. The search was performed in seven academic databases, resulting in 43 papers included in the main analysis. The results from the review depict the ongoing design patterns and learning phenomena that emerged from the synergy that learning analytics and learning design impose on the current status of learning technologies. Finally, this review stresses that future research should consider developing a framework on how to capture and systematize learning design data grounded in learning analytics and learning theory, and document what learning design choices made by educators influence subsequent learning activities and performances over time.

With the adoption of Learning Management Systems (LMSs) in educational institutions, a lot of data has become available describing students' online behavior. Many researchers have used these data to predict student performance. This has led to a rather diverse set of findings, possibly related to the diversity in courses and predictor variables extracted from the LMS, which makes it hard to draw general conclusions about the mechanisms underlying student performance. We first provide an overview of the theoretical arguments used in learning analytics research and the typical predictors that have been used in recent studies. We then analyze 17 blended courses with 4,989 students in a single institution using Moodle LMS, in which we predict student performance from LMS predictor variables as used in the literature and from in-between assessment grades, using both multi-level and standard regressions. Our analyses show that the results of predictive modeling, notwithstanding the fact that they are collected within a single institution, strongly vary across courses. Thus, the portability of the prediction models across courses is low. In addition, we show that for the purpose of early intervention or when in-between assessment grades are taken into account, LMS data are of little (additional) value. We outline the implications of our findings and emphasize the need to include more specific theoretical argumentation and additional data sources other than just the LMS data.

As Education constitutes an essential development standard for individuals and societies, researchers have been exploring the use of Artificial Intelligence (AI) in this domain and have embedded the technology within it through a myriad of applications. In order to provide a detailed overview of the efforts, this article pays particular attention to these developments by highlighting key application areas of data-driven AI in Education; it also analyzes existing tools, research trends, as well as limitations of the role data-driven AI can play in Education}. In particular, the article reviews various applications of AI in Education including student grading and assessments, student retention and drop-out predictions, sentiment analysis, intelligent tutoring, classroom monitoring and recommender systems. The article also provides detailed bibliometric analysis to highlight the salient research trends in AI in Education over seven years (2014–2020) and further provides detailed description of the tools and platforms developed as the outcome of research and development efforts in AI and Education. For the bibliometric analysis, articles from several top venues are analyzed to explore research trends in the domain. The analysis shows sufficient contribution in the domain from different parts of the world with a clear lead for the United States. Moreover, students' grading and evaluation have been observed as the most widely explored application. Despite the significant success, we observed several aspects of education where AI alone can not contribute much. We believe such detailed analysis is expected to provide a baseline for future research in the domain.

Although virtual reality (VR) simulation training has gained prominence, review studies to inform instructors and educators on the use of this technology in science, technology, engineering, and mathematics (STEM) are still scarce. This article presents various VR-supported instructional design practices in K-12 (primary and secondary) and higher education in terms of participants' characteristics, methodological features, and pedagogical uses in alignment with applications, technological equipment, and instructional design strategies. During the selection and screening process, 41 (n = 41) studies published in the period 2009-2019 were included for a detailed analysis and synthesis. This article's results indicate that many studies were focused on the description and evaluation of the appropriateness or the effectiveness of applied teaching practices with VR support. Several studies pointed out improvements in learning outcomes or achievements, positive perspectives on user experience, and perceived usability. Nevertheless, fewer studies were conducted to measure students' learning performance. The current scoping review aims to encourage instructional designers to develop innovative VR applications or integrate existing approaches in their teaching procedures. It will also inform researchers to conduct further research for an in-depth understanding of the educational benefits of immersive-VR applications in STEM fields.

Research has increasingly explored the application of artificial intelligence (AI) and machine learning (ML) within the mental health domain to enhance both patient care and healthcare provider efficiency. Given that mental health challenges frequently emerge during early adolescence– the critical years of high school and college– investigating AI/ML-driven mental health solutions within the education domain is of paramount importance. Nevertheless, conventional AI/ML techniques follow a centralized model training architecture, which poses privacy risks due to the need for transferring students' sensitive data from institutions, universities, and clinics to central servers. Federated learning (FL) has emerged as a solution to address these risks by enabling distributed model training while maintaining data privacy. Despite its potential, research on applying FL to analyze students' mental health remains limited. In this paper, we aim to address this limitation by proposing a roadmap for integrating FL into mental health data analysis within educational settings. We begin by providing an overview of mental health issues among students and reviewing existing studies where ML has been applied to address these challenges. Next, we examine broader applications of FL in the mental health domain to emphasize the lack of focus on educational contexts. Finally, we propose promising research directions focused on using FL to address mental health issues in the education sector, which entails discussing the synergies between the proposed directions with broader human-centered domains. By categorizing the proposed research directions into short- and long-term strategies and highlighting the unique challenges at each stage, we aim to encourage the development of privacy-conscious AI/ML-driven mental health solutions.

Knowledge Tracing (KT) is a core task in intelligent education. It aims to dynamically track students' knowledge mastery and predict their learning performance based on their historical answering records. Most existing mainstream methods focus on static feature enhancement while ignoring the interference of non-cognitive noises such as careless errors and random guesses. Moreover, current models cannot adaptively adjust cognitive representations according to learners' individual differences. Therefore, it is necessary to consider the expressiveness of cognitive representations from the perspectives of students' personalized answering performance and cognitive laws. To address these issues, this paper proposes an Adaptive Cognitive Representation Orchestration Knowledge Tracing Model (AOCR-KT). Specifically, we design an Adaptive Cognitive Optimization Module. This module integrates answering status and question difficulty, and introduces a it p-mechanism involving interval performance and continuity rules to compensate for the subjective bias and inaccuracy caused by revising the status only from the perspective of difficulty. The relevant parameters are fed into the adaptive dynamic programming algorithm to judge and revise the coherence and continuity of the answering status, gradually minimize the cost function, and drive the answering status toward the optimal solution. In addition, we construct a Partition Optimization Module. Considering the staged characteristics of learners' cognitive development, the answering sequence is divided into multiple intervals according to individual differences, and each interval is optimized independently before global optimization. Finally, relation embeddings generated via bipartite graph modeling are fused with the optimized cognitive representations to further enhance cognitive expression. Extensive experiments on three large-scale public educational datasets demonstrate that the AOCR-KT model outperforms state-of-the-art knowledge tracing models, which fully validates its effectiveness and superiority. The source code of AOCR-KT is available athttps://github.com/bigdata-graph/AOCR-KT.

Academic self-esteem (ASE) plays a significant role in children's learning outcomes by influencing their engagement and performance in subjects, such as math. Although existing studies emphasize the importance of self-esteem in academic achievement, the potential of interactive technologies, particularly virtual robots, in enhancing ASE remains underexplored. This study investigates the impact of a virtual robot on children's ASE, math performance, concentration, and engagement in an e-learning environment. The study involved an experimental group (n= 12) interacting with a virtual robot integrated into the math e-learning platform, and a control group (n= 12) working in a traditional e-learning setting without robot interaction. The results demonstrated that the experimental group, which interacted with the virtual robot, exhibited significant improvements in math performance, concentration, and engagement across the three experimental sessions (Sessions 1, 2, and 3) compared with the control group, as indicated by both quantitative measures and qualitative feedback from participants. ASE, as well as the quantity and quality of friendships, was assessed pre-and post-test, with findings indicating greater improvements in the experimental group after the intervention. The correlation between improved math performance and higher ASE was moderate to strong. These findings underscore the potential of virtual robots as tools that positively influence the ASE and learning outcomes of children, and highlight their value for future educational settings where such technology could address achievement gaps in mathematics learning through improved self-perception.

University graduates often encounter challenges during their initial job-seeking, such as interview anxiety and insufficient interview experience. To address these issues, this paper designs a job-seeking assistance system that integrates interview training and personalized job recommendation. The system employs Virtual Reality (VR) technology and Large Language Model (LLM) to create immersive interview scenarios for experiential learning. This approach provides a realistic interview simulation environment that facilitates the accumulation of interview experience and helps alleviate interview anxiety. Based on the interview dialogue records, we further propose a novel job recommendation with dual-view-enhanced hybrid expert module to improve recommendation accuracy. Experimental results on a real-world dataset demonstrate the optimal performance of the proposed model. After the interview, the system automatically generates interview analysis reports, including a competency analysis of the user, an evaluation of the target position, and a personalized job recommendation. Although the system is developed and evaluated within the context of the Chinese job market, it offers potential for extension to other cultural and labor market settings.

Blended learning enriches students' experiences across diverse environments while generating multimodal data related to learning activities. However, it presents challenges in the appropriate use of multimodal data to track students' performance development. Previous models with fixed-length inputs or static fusion mechanisms inadequately model temporal dependencies across behavioral modalities. In this paper, we integrated variable-length time series over weeks to forecast performance for the subsequent week. As the main contribution, a two-stage training model that relies on a transformer for temporal attention-based multimodal fusion was proposed. We conducted experiments on two real-world datasets, FC2023 and CS2023, derived from hybrid mode courses involving 439 and 199 students, respectively. The results demonstrate that multimodal fusion yields better periodical prediction compared to the unimodal approach. Ultimately, aiming at predicting the week-by-week development of student performance, the proposed model achieves the Area Under the Curve of Receiver Operating Characteristic (ROC-AUC) of 81.02% on FC2023 and 82.65% on CS2023. This approach, which leverages multimodal learning analytics, helps educators track each student's learning progress more effectively, enabling the timely implementation of instructional interventions and enhancing educational outcomes.

Non-anchor equating presents a significant challenge in educational assessment when test forms lack common items, requiring innovative solutions to ensure score comparability across different test administrations. This study proposes a novel Large Language Model-Simulated Non-equivalent Groups with Anchor Test (LLM-SNGAT) method that leverages large language models to simulate test-taking samples and generate common item sets for equating purposes. The approach eliminates traditional dependencies on specialized test design and extensive demographic data col-lection by utilizing the inherent capabilities of large language models to simulate diverse response patterns. We evaluated the method using Tucker and Levine equating approaches across multiple LLMs, including GPT-4o, O1-preview and DeepSeek-R1. Results demonstrated the feasibility of the proposed approach, with the Tucker method showing superior performance and consistent improvements as common item coverage increased. Sensitivity analysis confirmed that model performance rankings remained consistent across varying prompt formulations. The study revealed characteristic that standard errors were smallest near the mean and became larger farther away from the mean, and identified optimal common item proportions of 30-50% for stable equating performance. While current limitations include the capacity of large language models to accurately simulate human cognitive and behavioral diversity, this proof-of-concept study provides preliminary evidence for the feasibility of the LLM-SNGAT methodology. The approach represents a paradigm shift from resource-intensive traditional methods to computationally-driven solutions, offering promising prospects for addressing non-anchor equating challenges in the digital age.

Aesthetic perception, the cognitive process through which individuals interpret and evaluate the expressive and emotional qualities of visual art, is fundamental to students' creative and emotional development. Recent progress in artificial intelligence has enabled computational models to assist in aesthetic analysis by identifying patterns in visual composition and affective expression. However, such models often struggle to recognize abstract or context-dependent aesthetic dimensions, and improving these aspects through comprehensive annotation remains costly and time-consuming. This study presents a human-AI teaming framework designed to identify the aesthetic perception dimensions that AI models find most difficult to interpret and to allocate these dimensions to human experts for annotation. The framework employs a multi-agent reinforcement learning (MARL) mechanism, where each agent is assigned to a specific aesthetic dimension and learns a policy for determining whether expert annotation is required. Two complementary state representation strategies are introduced: a statistical representation that captures the model's predictive distribution across dimensions, and a graph-based attention module that models interdependencies among aesthetic attributes. A reward mechanism further guides agents to balance the improvement of model perception with the minimization of human annotation effort. Experiments conducted on two real-world datasets demonstrate that the proposed framework effectively identifies the challenging dimensions for AI models and strategically delegates them for human evaluation. This targeted collaboration significantly enhances annotation efficiency and model interpretability, providing a scalable approach for improving human-AI synergy in aesthetic perception analysis.

Automated coding is particularly crucial in online learning research, where vast amounts of text provide valuable insights into cognitive engagement, emotional expression, and social interaction, yet manual analysis of such large-scale discourse remains time-consuming. While Large Language Models (LLMs) offer promising solutions, existing approaches suffer from inconsistent coding performance and suboptimal prompt engineering, limiting their reliability across diverse educational frameworks. This study introduces the Optimized Large Language Model Coding (OLLM-C) framework, a systematic approach that enhances automated content analysis through calibrated prompting strategies. We evaluated OLLM-C using a comprehensive dataset of 8,671 comments collected from five online courses. The framework was validated against 15 widely-used coding frameworks spanning cognitive, social-emotional, and behavioral dimensions. Comparative analysis of five LLMs (GPT-3.5, GPT-4o, Gemini, Claude, and Llama) revealed GPT-4o's superior performance in initial coding tasks. The OLLM-C calibration process significantly improved GPT-4o's reliability, with Cohen's Kappa coefficients increasing from an average of 0.45 to 0.57 across frameworks. Notably, the model demonstrated stronger performance in social-emotional coding compared to cognitive skills frameworks, achieving substantial agreement with human coders in emotion recognition and social interaction analysis while showing limitations in complex cognitive reasoning tasks. These findings establish OLLM-C as a systematic calibration framework that enhances the reliability, efficiency, and practical applicability of LLM-assisted qualitative analysis.